9898
J. Phys. Chem. B 1998, 102, 9898-9907
Quantized Capacitance Charging of Monolayer-Protected Au Clusters Shaowei Chen and Royce W. Murray* Kenan Laboratories of Chemistry, UniVersity of North Carolina, Chapel Hill, North Carolina 27599-3290
Stephen W. Feldberg* Department of Applied Sciences, BrookhaVen National Laboratory, Upton, New York 11973-5000 ReceiVed: June 30, 1998; In Final Form: July 31, 1998
A theoretical model based on electrostatic interactions is developed to account for the formal potentials of current peaks observed in differential pulse voltammetry of solutions of 10 different nanometer-sized alkylthiolate and arylthiolate monolayer-protected gold clusters. The current peaks arise from successive, quantized (single-electron) capacitative charging of ensembles of individual cluster cores (i.e., electrochemical ensemble Coulomb staircase charging). Experimental peak potentials for a given cluster change roughly linearly with changes in its core charge state, as predicted by the theory, and the sub-attofarad capacitances (CCLU) of individual clusters obtained from the slopes of such plots agree with those estimated from a simple concentric sphere capacitor model. The charging of clusters with very small cores becomes redox moleculelike, indicating as reported recently, the emergence of HOMO-LUMO energy gaps. Because the quantized charging currents of the clusters are diffusion controlled, their voltammetric behavior can be readily simulated, but requires attention to dispersities in CCLU that occur in experimental samples of these materials. Simulations of microelectrode voltammetry incorporating Gaussian dispersions in cluster properties display features similar to those observed experimentally. The simulations show that quantized charging features are more difficult to detect when the nanoparticles are not monodisperse, but can be seen in polydisperse samples when the cores are small (small CCLU) and not highly charged.
Introduction An extensive research interest in nanometer-sized metal and semiconductor particles (“nanoparticles”) has arisen because of fundamental questions about this largely unexplored materials dimension and because of its diverse applications in optoelectronic devices, molecular catalysts, and chemical sensors.1,2 Despite extensive preparatory studies,2 it has only recently become possible, starting with a report by Brust et al.,3 to isolate nanometer-sized metal (Au) clusters in solvent-free form that are stable and do not aggregate, can be redissolved without change, and can even be subjected to further synthetic manipulations.4 Stabilization of the Au (and subsequently described Ag,5 Pt,6 and alloy7) clusters is accomplished by protection by ligand monolayers of alkanethiolates and other thiolate ligands.8-10 The stability of these monolayer-protected clusters (MPCs) has made possible their solubility-based fractionation into monodisperse populations of nanoparticles.11 Because of a combination of small metal-like core size and hydrocarbon-like dielectric coating, the capacitance (CCLU) of an MPC can be less than an attofarad (aF) per MPC in (toluene/acetonitrile) electrolyte solutions. Addition or removal of single electrons from such tiny capacitors produces potential changes [e/CCLU] . kBT/e, so that the solution phase, double-layer capacitance charging of the MPCs becomes a quantized property. This property of MPCs was demonstrated recently12 and is an electrochemical analogue to Coulomb staircase charging.13 A further study has revealed a transition from metal-like double-layer capacitance charging to molecule-like redox charging as the MPC core size was decreased.14 We have since additionally observed quantized capacitance charging for arylthiolate-MPCs.15 Using differential
pulse voltammetry (DPV) to observe the quantized charging, data are now in hand for 10 cluster solutions which have displayed multiple-peak DPV responses. Sequential electron transfers observed with fullerene (and derivatives)16 and Pt-carbonyl nanoclusters [Ptn(CO)m]17 led Weaver et al.18 to develop an electrostatic model to relate the electron-transfer energetics of molecular capacitances in gasand solution-phase systems. This paper uses this earlier theory to explain the charging energetics of Au MPCs and expands it to the diffusion-controlled voltammetry of monodisperse and polydisperse (in core size) Au MPCs. The Au MPCs differ from the fullerenes and Pt-carbonyl clusters in that they (a) display oxidative as well as reductive charging, (b) are somewhat larger and tend (except for the smallest examples14) to be metallike, and (c) involve a monolayer shell the thickness and chemical dielectric characteristics of which can, in principle, be systematically varied. The general conclusion drawn from the electrostatic theory in light of the available charging data is that the theory quite well represents the potentials of DPV peaks for successive, single-electron charging of MPCs. The “formal potentials” of the charging peaks are referenced to the potential of zero charge of the clusters (EPZC, which is independently established). Exceptions to the predicted electrostatic behavior are seen for the smallest and most molecule-like of the MPCs, owing to the development of HOMO-LUMO gaps, and are manifested in large spacings between charging peaks immediately adjacent to the EPZC. We also observe that since currents for kinetically rapid quantized charging steps are controlled by diffusion of the MPCs to the electrode,12,14 their voltammetry can be
10.1021/jp982822c CCC: $15.00 © 1998 American Chemical Society Published on Web 10/21/1998
Monolayer-Protected Au Clusters
J. Phys. Chem. B, Vol. 102, No. 49, 1998 9899
described using the classical combination of Nernst and Fick’s equations. We illustrate this using simulations of steady-state microelectrode voltammograms. The simulations also show that dispersity in the MPC capacitance causes gradual loss of quantized charging features in the voltammetry, just as observed experimentally.15
Theory The theoretical approach to formulating the potentials at which successive single electron transfers arising from quantized charging of Au MPCs occur begins with simple electrostatic interactions, as described by Weaver et al.18 We extend his theory to point out the Nernstian behavior of such charging (assuming that the electron-transfer dynamics between the working electrode and MPCs are fast), to simulate the corresponding microelectrode voltammetry, and to account for effects of capacitance dispersity. Simplifications of the theory are (a) the MPC capacitance is based on an assumed spherical Au core (the actual core shape is postulated to be a truncated octahedron (TO)8 or truncated decahedron (TD)19), (b) effects of HOMOLUMO gaps are neglected, and (c) the MPC capacitance is assumed to be independent of its state of charge. Formal Potentials of Quantized Charging Reactions. This section describes the distribution of charges in a solution of monodisperse, spherical, metal-like nanoparticles (e.g., MPCs) in equilibrium with a working electrode. Of particular interest are the effects of successive single electron charging of the MPC, namely, the potentials at which these steps occur. The basic premise of the analysis is that the distribution of electrons between the working electrode and the MPCs is determined by the potential applied to the working electrode (EAPP), the potential of the MPC (EP), and the (integral) capacitance (CCLU) of an individual MPC:
EP ) EPZC +
ze CCLU
(1)
where EPZC is the potential of zero charge of the nanoparticle, z is the number of electronic charges on the particle and is signed (i.e., if z > 0, the particle is oxidized while if z < 0, reduced), and e is the electronic charge (1.602 × 10-19 C). The capacitance (CCLU) of a spherical nanoparticle coated with a uniform monolayer (of dielectric constant ) and in an electrolyte solution can be expressed as
CCLU ) ACLU
0 r + d r ) 4π0 (r + d) r d d
(2)
where E0 is the permittivity of free space, ACLU the surface area of the MPC core of radius r, and d the thickness of the protecting dielectric monolayer. Equation 2 predicts that when core size and protecting monolayer thickness are comparable (in the present study, r ) 0.5-1.5 nm and d ) 0.5-1.0 nm), CCLU increases with increasing core radius, and for monolayers with similar dielectric property, decreases with increasing monolayer thickness. Considering the limits of eq 2, when r , d, CCLU ) 0ACLU/r, (e.g., a naked nanoparticle in a thick dielectric medium) and when r . d, CCLU ) 0ACLU/d (akin to a monolayer on a flat surface). The work W required to charge initially uncharged MPCs by z electrons to a potential EP (see eq 1) with a working electrode
at potential EAPP is (at equilibrium)
W)
[
∫0z e(EAPP - EP) dz ) ez EAPP -
(
EPZC +
)]
ze 2CCLU
(3)
At the electrode/solution interface, the corresponding Boltzmann population (NZ) of MPCs having z charges relative to the population of uncharged MPCs (N0) is
( )
{ [
(
)]}
NZ W ze ze ) exp ) exp E - EPZC + N0 kBT kBT APP 2CCLU
(4)
where kB is the Boltzmann constant (1.38 × 10-23 J/K) and T is absolute temperature (K). The ratio (Rz) of the populations of adjacent charged states (differing by one electron in charge), NZ/NZ-1, is easily deduced from eq 4 as
RZ )
{ [
]}
NZ (z - 1/2)e e ) exp EAPP - EPZC NZ-1 kBT CCLU
(5)
An important observation at this point is that eq 5 is Nernstian in its form. Thus, a mixture of MPC particles having charge z and z-1 in a solution comprise, in a formal sense, a mixed valent solution of a “redox couple” with a formal potential, EoZ,Z-1, which is o ) EPZC + EZ,Z-1
(z - 1/2)e CCLU
(6)
The substance of this analogy of redox and nanoparticle behavior is that the latter can be formally regarded as a multivalent redox system which exhibits equally spaced formal potentials (assuming that CCLU is independent of the charge state of the MPC). A major difference, of course, between nanoparticles and conventional multivalent redox systems (such as [Ru(bpy)3]3+/2+/1+/0 where bpy ) 2,2′bipyridine), is that the electrochemical formal potentials of the latter are generally not evenly spaced for the first versus subsequent electron transfers, owing to molecular features of the complexes such as electronic coupling, ligand-metal or metal-metal interactions, and HOMO-LUMO gaps.20 When such molecular effects are absent, that is, when CCLU is independent of z, eq 5 predicts that the formal potentials of the capacitance charging steps should vary linearly with the valence states of the nanoparticle. From such plots one can evaluate the aVerage nanoparticle capacitance. Additional information is required to identify the EPZC and the adjacent charging steps for which z ) (1/0. Core Size Distribution. Equation 2 shows that the capacitance CCLU of the MPC depends on both core radius and protecting monolayer thickness. Experimentally, while the core shape cannot be accurately discerned, the core size can be evaluated by transmission electron microscopy (TEM). In addition, there is always some degree of dispersity of the observed core size, and therefore some dispersity of the cluster capacitance. Cluster samples as-prepared typically display some polydispersity, but repeated fractional recrystallization11 can produce MPCs with a narrower size distribution and hence a more uniform value of CCLU (i.e., approaching ideal monodispersity). (It is worth observing that “dispersity” is not a simple concept and in fact requires consideration of variations in both core and monolayer. Also, experimental methods will not be equally sensitive to all forms of dispersity. That is, TEM and mass spectrometry detect core size dispersity. Electrochemical
9900 J. Phys. Chem. B, Vol. 102, No. 49, 1998
Chen et al.
Figure 1. (A) TEM core size distribution (bar) of C4Au (28 kD) clusters and fit (s) with bi-Gaussian equation (eq 7b): x1 ) 0.14, r01 ) 0.63 nm, σ1 ) 0.01 nm, x2 ) 0.86, r02 ) 0.74 nm, σ2 ) 0.34 nm. (B) Computed bi-Gaussian core size distribution with varying compositions of two populations using the above r0 and σ values (the maxima of the smaller core distributions have been truncated).
experiments detect dispersity in MPC capacitance, which can be determined by a combination of variation in the core size and packing in the monolayer shell, and by variations in the number of thiolate ligands attached to each MPC core. Fractional recrystallization separates MPCs on the basis of solubility dispersity, which is determined by a complex combination of all the monolayer structural and core size variations. In the present theory, we consider the dispersity in MPC capacitance as arising from variations on the core radius r. Considering dispersity alternatively as variations in the dielectric monolayer thickness d involves quite similar computations, which are not presented here.) Clusters that have been fractionally recrystallized, and sometimes clusters that are crude synthetic products (i.e., arylthiolated Au MPCs15), give TEM images and core size distributions that can be approximated by a single
f(r) )
or a bi-Gaussin function, i.e.,
f(r) )
{ [
[
]
(r - r0)2 1 exp σ2 σxπ
]
(r - r01)2 1 x exp + σ21 xπ σ1
(7a)
Electrochemical Responses. We have detected quantized charging of MPCs in solutions by several electrochemical techniques including pulse, cyclic, and steady-state microelectrode voltammetry.12,14,15 The voltammetric responses are diffusion controlled. Steady-state microelectrode voltammetry is the simplest of these techniques, and the derivative of the current-potential response should approximate the shape of differential pulse voltammograms (DPV, with which most of the quantized charging peak potentials have been measured). The limiting current for steady-state microelectrode voltammetry is
i ) 4nrELFDC*
where n is the number of electrons transferred, rEL the microelectrode radius, F Faraday’s constant, D the MPC diffusion coefficient, and C* the MPC bulk concentration. The diffusion coefficient (D) is size-dependent and well approximated by the Einstein-Stokes equation
D)
[
(r - r02) 1-x exp σ2 σ2
]}
2
2
(7b)
where r0 is mean core radius and σ its standard deviation, subscripts 1 and 2 denote two populations with r01 and r02 mean core radii, and x is the mole fraction of the r01 population. As an example, Figure 1A shows the TEM-derived core size distribution of a butanethiolate-protected Au MPC (abbreviated C4Au) that was partially fractionated and measured (by laser ionization/desorption mass spectrometry) to have a mean core mass of ca. 28 kD.14 The core size (r) distribution of this sample can be roughly represented by a bi-Gaussian function (eq 7b) in which 14% of the MPC population is narrowly centered at radius r01 ) 0.63 nm (dispersity σ1 ) 0.01 nm) while a larger population (86%) is broadly distributed around radius r02 ) 0.74 nm (dispersity σ2 ) 0.34 nm). Figure 1B illustrates how the core size dispersity would change with various proportions of the two different populations of r01 and r02 values and associated dispersities.
(8)
kBT 6πηrH
(9)
where η is solvent viscosity and rH the hydrodynamic radius of the species. Results from rotating disk voltammetry21 suggest that MPC hydrodynamic radii can be roughly approximated by rH ≈ r + d. Combining eqs 7-9 gives the microelectrode current response as
i ) 4nrELF
∫rr
HIGH
(
f(r)
LOW
-1
0
Rz
N
+∑ ∑ -N 1 + R 1 1+R z
)
×
z
kBT
C* dr (10)
6πη(r + d) or
iNORM )
∫r
rHIGH
LOW
(
f(r)
0
-1
N
Rz
+∑ ∑ -N 1 + R 1 1+R z
)
z
dr
r+d
(11)
where iNORM is the normalized current [iNORM ) (6iπη)/ (4nrELFC*kBT)] and rLOW and rHIGH are, respectively, the lower and upper limits of the core sizes (as observed experimentally). In eqs 10 and 11, the first and second summation terms within
Monolayer-Protected Au Clusters
J. Phys. Chem. B, Vol. 102, No. 49, 1998 9901
TABLE 1: Formal Potentials of Quantized (Electrochemical Coulomb Staircase) Charging of Various Monolayer-Protected Au Nanoclustersa Au MPCs [core radius] (nm)b C4Au (14 kD) [0.65] C4Au (22 kD) [0.71] C4Au (28 kD) [0.81]
C6Au (8 kD) [0.55] C6Au (22 kD) [0.71] C6Au (28 kD) [0.81]
C6Au (38 kD) [0.95]
PhC4SAu [1.05]
PhC2SAu [1.06]
4-CresolSAu [1.50]
z/z-1c
EP,C (V)d
EP,A (V)d
EP (V)d
∆EP (V)d
+2/+1 +1/0 0/-1 -1/-2 +2/+1 +1/0 0/-1 -1/-2 -2/-3 +4/+3 +3/+2 +2/+1 +1/0 0/-1 -1/-2 +2/+1 +1/0 0/-1 -1/-2 +2/+1 +1/0 0/-1 -1/-2 -2/-3 +5/+4 +4/+3 +3/+2 +2/+1 +1/0 0/-1 -1/-2 -2/-3 +5/+4 +4/+3 +3/+2 +2/+1 +1/0 0/-1 -1/-2 -2/-3 +3/+2 +2/+1 +1/0 0/-1 -1/-2 -2/-3 -3/-4 +7/+6 +6/+5 +5/+4 +4/+3 +3/+2 +2/+1 +1/0 0/-1 -1/-2 -2/-3 +4/+3 +3/+2 +2/+1 +1/0
0.530 0.334 -0.410 -0.714 0.438 0.144 -0.255 -0.604 -1.000 0.714 0.550 0.314 0.010 -0.298 -0.600 0.670 0.370 -0.850 -1.080 0.270 -0.120 -0.420 -0.760 -1.000 0.933 0.717 0.558 0.308 -0.095 -0.376 -0.688 -1.031 0.830 0.640 0.430 0.120 -0.200 -0.430 -0.690 -0.850 0.664 0.447 0.054 -0.180 -0.394 -0.926 -1.310 1.000 0.800 0.600 0.458 0.310 0.164 -0.012 -0.206 -0.383 -0.612 0.462 0.242 0.086 -0.046
0.582 0.304 -0.428 -0.700 0.483 0.175 -0.235 -0.517 -0.947 0.709 0.520 0.278 -0.056 -0.332 -0.632 0.820 0.490 -0.740 -0.920 0.330 -0.060 -0.360 -0.680 -0.910 0.894 0.736 0.552 0.288 -0.096 -0.392 -0.704 -0.976 0.970 0.830 0.580 0.210 -0.076 -0.380 -0.550 -0.740 0.795 0.424 0.060 -0.192 -0.418 -0.924 -1.260 1.000 0.900 0.642 0.483 0.340 0.184 0 -0.204 -0.400 -0.512 0.400 0.216 0.068 -0.088
0.556 0.319 -0.419 -0.707 0.461 0.160 -0.245 -0.561 -0.974 0.712 0.535 0.296 -0.023 -0.315 -0.616 0.745 0.430 -0.795 -1.000 0.300 -0.090 -0.390 -0.720 -0.955 0.914 0.727 0.555 0.298 -0.096 -0.384 -0.696 -1.004 0.900 0.735 0.505 0.165 -0.138 -0.405 -0.620 -0.795 0.730 0.436 0.057 -0.186 -0.406 -0.925 -1.285 1.000 0.850 0.621 0.471 0.325 0.174 -0.006 -0.205 -0.392 -0.562 0.431 0.229 0.077 -0.067
0.052 0.030 0.018 0.014 0.045 0.031 0.020 0.087 0.053 0.005 0.030 0.036 0.066 0.034 0.032 0.150 0.120 0.110 0.160 0.060 0.060 0.060 0.080 0.090 0.039 0.019 0.006 0.020 0.001 0.016 0.016 0.055 0.140 0.190 0.150 0.090 0.124 0.050 0.140 0.110 0.131 0.025 0.006 0.012 0.024 0.002 0.050 0 0.100 0.042 0.025 0.030 0.020 0.012 0.002 0.017 0.100 0.062 0.026 0.018 0.042
∆VC (V)d 0.237 0.738 0.288 0.301 0.405 0.316 0.413 0.177 0.239 0.319 0.292 0.301 0.315 1.225 0.205 0.390 0.300 0.330 0.235 0.187 0.172 0.257 0.394 0.288 0.312 0.308 0.165 0.230 0.340 0.303 0.267 0.215 0.175 0.294 0.379 0.243 0.220 0.519 0.360 0.150 0.229 0.150 0.146 0.151 0.180 0.199 0.187 0.170 0.202 0.152 0.144
a Peak positions were determined by differential pulse voltammetry (DPV) (experimental details in refs 14 and 15). While all were determined vs Ag/AgCl reference, differences in peak positions are experimentally much more certain than absolute potentials. b C4Au and C6Au denote butanethiolate and hexanethiolate MPCs, respectively; and PhC4SAu, 4-phenylbutyl-1-thiolate MPC; PhC2SAu, 2-phenylethylthiolate MPC; 4-CresolSAu, 4-thiolcresol MPC (these three arylthiolate MPCs were prepared at 0 °C and with 3-fold excess of thiol over Au) (ref 15); the core radii given for these MPCs are determined by TEM measurements (refs 10 and 15). c Valence state changes based on double-step chronocoulometric studies (ref 14), solution rest potentials (ca. -0.1 V), and impedance spectroscopic studies of nanoparticle monolayers on Au electrode surfaces where the potential of zero charge is ca. -0.2 V (ref 15). d Subscripts (A and C) denote anodic and cathodic peaks, EP ) (EP,A + EP,C)/2, ∆EP ) EP,A - EP,C, ∆VC is the peak spacing between two adjacent peaks.
9902 J. Phys. Chem. B, Vol. 102, No. 49, 1998
Chen et al.
the parenthesis represent current contributions from adding electrons to or removing electrons from the MPC in the charging step, respectively. These equations can be used for numerical simulations and/or data fitting. Simulations of cyclic voltammetry (CV) for macroelectrodes, where currents are not steady state, should be equally straightforward but are less valuable for comparison to experiment, since MPCs tend to physisorb from their solutions onto electrode surfaces with a coverage depending on the cluster concentration.22 In CV experiments, the resulting currents can be a combination of reactions of adsorbed and diffusing MPCs. Nonetheless, if the potentials for the successive charging steps are evident in the CV responses, they can be straightforwardly compared to the predictions of eqs 2 and 6. Simulations and Discussion In this section, the theory for MPC capacitance is compared to experimental values, followed by simulations of the steadystate microelectrode voltammetry of Au MPCs of different core sizes, protecting thiolate monolayers, and dispersity. MPC Capacitances and Formal Potentials. Table 1 lists experimental results for differential pulse voltammetry (DPV) peak potentials, observed between -1 to +1 V vs Ag/AgCl for electrolyte solutions (mixed toluene/acetonitrile solvent) of 10 different Au MPCs (details of MPC nomenclature in Table 1 footnotes). Each peak potential represents a discernible charging step (above background). Experimental conditions and examples of DPV current-potential curves from which these data are derived have been previously presented.12,14 The small differences (∆Ep) between DPV peak potentials in negativeand positive-going potential scans are mostly uncompensated iR effects. Small substructures (i.e., shoulders) occasionally seen in the DPV current-potential curves are ignored at this level of analysis. The parameter ∆VC represents the difference between adjacent charging peaks. All of the alkanethiolateMPC samples and the PhC4SAu sample in Table 1 had been fractionated; the other two arylthiolate-MPCs were unfractionated. The assignment shown for the valence state change (Z/Z-1) for each peak in Table 1 is based on EPZC having a value near -0.1 to -0.2 V (vs Ag/AgCl). This assignment was established by ac impedance measurements on a monolayer of C4Au (28 kD) MPCs attached to a Au electrode surface, that show a shallow double-layer capacitance minimum at ca. -0.2 V (vs Ag/AgCl)15 (i.e., a PZC around -0.2 V). This surface should be a good model for an MPC in solution. Additionally, rest potentials of naked Au electrodes in MPC solutions in this solvent are generally around -0.1 V. Thus, their reduction and oxidation processes occur at potentials more negative or positive than about -0.2 V, respectively. Figure 2 compares the MPC formal potential (EP) results from DPV to eq 6. Overall, the linearities of the changes of formal potential with valence state are excellent. (The exceptions to good linear fitting, the two smallest MPCs, 14 and 8 kD in Figures 2A and B, respectively, are discussed further below.) The overall Figure 2 result confirms our previous assertion14 that, for larger MPCs, the quantized capacitance charging steps can be regarded as double layer capacitance phenomena. The quantized capacitance charging in Table 1 and Figure 2 is analogous to Coulomb staircase charging of single nanoparticles, where equally spaced charging steps are predicted by theory and are experimentally seen. The Figure 2 slopes and intercepts give corresponding MPC capacitances (CCLU,EQN6) and EPZC (eq 6), in Table 2. Table 2
Figure 2. Formal potentials of the quantized capacitance charging of various Au MPCs versus their valence states where z refers to the couple z/z-1 (eq 6). Lines shown are linear regression.
TABLE 2: Formal Potentials of Quantized Capacitance Charging of Various Au Nanoparticles Au MPCs
CCLU∆V (aF)a
CCLU,EQN2 (aF)b
CCLU,EQN6 (aF)c
EPZC (V)c
C4Au (14kD) C4Au (22kD) C4Au (28kD) C6Au (8kD) C6Au (22kD) C6Au (28kD) C6Au (38kD) PhC4SAu PhC2SAu 4-CresolSAu
0.22 0.40 0.50 0.13 0.41 0.41 0.53 0.66 0.89 1.11
0.49 0.56 0.69 0.31 0.45 0.55 0.71 0.74 1.22 2.79
0.35 0.44 0.59 0.25 0.51 0.56 0.62 0.49 0.93 0.97
-0.06 -0.05 -0.17 -0.16 -0.21 -0.24 -0.22 -0.06 -0.12 -0.16
a From quantized capacitance charging equation C CLU∆V ) e/∆VC where ∆VC is the difference between peaks for z ) (1/0 [refs 12, 14, and 15]. b From eq 2: for C4 and C6 monolayers, ) 3, dC4 ) 0.52 nm, and dC6 ) 0.77 nm; for PhC4S monolayer ) 3, d ) 0.94 nm, r ) 1.1 nm; for PhC2S monolayer ) 4, d ) 0.67 nm, r ) 1.1 nm; for 4-CresolS monolayer ) 5, d ) 0.64 nm, r ) 1.5 nm; monolayer thicknesses (d) are approximated as fully extended chainlengths of the corresponding ligands (Hyperchem). c From plots of eq 6 (Figure 2).
also gives MPC capacitances (CCLU∆V) derived (using the simple relation (VC ) e/CCLU∆V) from the spacings between DPV peaks that are adjacent to EPZC (i.e., z ) (1/0). (The CCLU∆V results were reported previously14 for the alkanethiolate-MPCs.) Also
Monolayer-Protected Au Clusters
J. Phys. Chem. B, Vol. 102, No. 49, 1998 9903
Figure 3. (A) Formal potentials of and (B) peak spacing between the z ) (1/0 peaks of C6:Au MPCs with various core sizes. (b) z ) +1/0, and (O) z ) 0/-1 (Data points of r ) 0.65 nm are for C4Au MPC, 22 kD). The lines (s) and (‚‚‚) are only guides to the eye. Vertical bars in part A are the theoretical predictions of ∆VC for the C6:Au MPCs based on their CCLU,EQN2 values in Table 2, centered around a supposed EPZC of -0.2V (- ‚ ‚ -). The values of ∆VC for illustrative larger particle dimensions (the “Bulk” domain) correspond to calculations from eq 2 using ) 3, dC6 ) 0.77 nm, and r ) 5 nm (a colloidal particle) and r ) 5 µm (a small microelectrode disk), giving 13 mV and 0.37 µV, respectively.
given in Table 2 are capacitances CCLU,EQN2 calculated from eq 2, using reasonable estimates of the relevant parameters (see table footnotes). Table 2 results for EPZC lie between -0.05 and -0.22 V, which is consistent with how the z ) (1/0 valence state changes were assigned (vide supra). Variation of EPZC values among the MPCs may reflect some structural properties, but equally likely, represent experimental uncertainty. It has been observed before for alkanethiol monolayers on flat surfaces23 that EPZC is difficult to determine. Examination of Table 2 shows that MPC capacitances estimated from the spherical core model (eq 2, CCLU,EQN2), agree reasonably well with those determined from plots of eq 6 (CCLU,EQN6). The former tend to be somewhat larger than CCLU,EQN6 values. On the whole, the spherical capacitor model of eq 2 works rather well. (Equation 2 produces a value for 4-CresolSAu MPCs that is much larger than the experimentally based capacitance. However, this MPC sample is unfractionated, and the core radius employed in the calculation (1.5 nm) is the average, TEM-derived value while the DPV peaks observed for this MPC are almost certainly dominated by the smaller nanoparticles in the distribution. The discrepancy thus reflects the less well-defined dimensions of the 4-CresolSAu MPCs.) Further examination of Table 2 shows that, for MPCs with core mass g22 kD, reasonably good agreement is found between the average capacitances determined from Figure 2 (CCLU,EQN6) and those (CCLU∆V) determined from the potential spacings between z ) 0/+1 and z ) 0/-1 chargings in DPV.12,14 For smaller MPCs (e14 kD), the larger difference that appears between the capacitances is attributed to the fact that, at sufficiently small size, the clusters develop molecule-like redox behavior (i.e., a HOMO-LUMO gap as ascertained by comparison of their electrochemistry and spectroscopy).14 The redox property produces a larger potential spacing between the z ) +1/0 and z ) 0/-1 MPC valence changes, and thereby a CCLU∆V capacitance that is smaller than that resulting from inclusion of charging peaks at more negative and positive potentials (according to eq 6, in the Figure 2 plots). Weaver et al.18 introduced an implicit quantum factor into his electrostatic relations, and the 8 and 14 kD MPCs of Table 2 are experimental examples which reflect such effects.
Figure 3 shows the effect of MPC core size on the charging potentials (part A) and the spacing between them (part B), for the Z ) +1/0 and 0/-1 peaks of C6Au MPCs. The z ) +1/0 and z ) 0/-1 peaks represent, formally, the solution analogues of ionization potential (IP) and electron affinity (EA), respectively.18 The enlarged spacing between the z ) +1/0 and z ) 0/-1 peaks for the two smallest core sizes in Figure 3 is consistent with the emergence of a HOMO-LUMO gap seen in near-IR spectroscopy of the clusters,14 and is greater than that anticipated with change in core radius from eq 2. The vertical bars in Figure 3A represent the predicted core radius dependency of ∆VC (calculated from CCLU,EQN2 in Table 2). Comparison of these theoretical predictions with the experimental points shows good agreement for MPCs with core radius ≈0.7 nm and mass g22 kD, but a substantial disagreement between the experimental EP spacings and predictions of ∆VC for the 8 and 14 kD core mass clusters. Figure 3B represents the changes in spacing in a way suggesting that the transition to more redox-like behavior14 begins at a cutoff core size of approximately 22 kD. (One would obviously like to see future additional experimental examples, on this important point of core size at which molecule-like properties start to occur.) Finally, to illustrate a transition to the more conventional domain of double layer charging of larger gold particles, predictions of ∆VC are shown at the right-hand side (large r) of Figure 3A for a small Au colloid particle (r ) 5 nm) and a small microelectrode. The predicted ∆VC values for these larger sizes are below the range of current (room temperature) observability. According to the preceding discussion, the radius domains of Figure 3A are thus labeled (roughly) in terms of observing charging in “redox”, “quantized” (metal-like), and “bulk” (metal) size domains. Figure 3A thus shows that metallike quantized capacitance charging can be expected to be observed only over a modest range of core dimensions. The above evaluation is based on the cluster capacitances being independent of charge states (i.e., CCLU is the same whether the MPC is positively or negatively charged). Some of the Table 1 data suggest that peaks are more closely spaced at extremes of positive and negative potentials, and it is wellknown that double-layer capacitance varies with potential.23 The Table 2 CCLU,EQN6 results average such effects. The effect of the sign of the MPC charge state was examined by plotting
9904 J. Phys. Chem. B, Vol. 102, No. 49, 1998
Chen et al.
TABLE 3: Capacitance and Potential of Zero Charge of Various Monolayer-Protected Au Nanoclusters (MPCs) Determined from Plots of Equation 6 for Electrochemical Peaks at the Positive and Negative Valent States Au MPCs
CCLU,EQN6+ (aF)a
CCLU,EQN6(aF)a
EPZC,+ (V)a
EPZC,(V)a
∆EPZC (V)b
C4Au (14kD) C4Au (22kD) C4Au (28kD) C6Au (8kD) C6Au (22kD) C6Au (28kD) C6Au (38kD) PhC4SAu PhC2SAu 4-CresolSAu
0.67 0.53 0.65 0.51 0.41 0.65 0.60 0.47 0.96 0.97
0.56 0.44 0.53 0.78 0.57 0.52 0.82 0.42 0.90
0.20 0.01 -0.11 0.27 -0.29 -0.13 -0.23 -0.10 -0.09 -0.16
-0.28 -0.05 -0.17 -0.69 -0.27 -0.23 -0.31 0.06 -0.12
0.48 0.06 0.06 0.96 0.02 0.10 0.08 0.04 0.03
a Subscripts (+ and -) denote states where z > 0 and z e 0, respectively. b (B) ∆EPZC ) |EPZC+ - EPZC-|.
data for the positive (z > 0) and negative (z e 0) valence states separately according to eq 6, again evaluating respectively (Table 3), values for CCLU,EQN6,+ (from slope) and CCLU,EQN6and EPZC+ and EPZC- (from intercepts). Overall, the CCLU,EQN6,+ and CCLU,EQN6- values for the various clusters are comparable and there seems to be no general bias that one is larger than the other. We conclude that the MPC capacitances are not very sensitive to the sign of their charge state (although surely they are not completely insensitive). The difference ∆EPZC between EPZC+ and EPZC- is more interesting. ∆EPZC is less than 0.1V for the larger core size clusters but is large for the two smallest MPCs (e.g., ca. 0.97 V for C6Au (8 kD) and 0.48 V for C4Au (14 kD)). This is yet another manisfestation of the onset14 of molecular-like redox properties for these materials. Simulations of Microelectrode Voltammetry. The above analyses demonstrate that, except for those with very small cores, the charging energetics of MPCs are well represented by simple electrostatic relations. Simulations of microelectrode voltammetry were carried out with eq 11, with emphasis on the effects of core dimensions, core size dispersity and monolayer characteristics. Figure 4 shows simulated voltammetry for solutions of completely monodisperse clusters of varied core sizes but fixed monolayer thickness. It is readily evident that the charging steps are most easily observed for the smaller core clusters (r < 1.25 nm). Constant current step heights and step
spacings are observed because the MPC capacitance is assumed to be independent of electrode potential, and the peak spacing decreases with increasing core size since ∆Vc ) e/CCLU (as in classical Coulomb staircases13). For sufficiently large clusters, the responses gradually become ill-defined, as the peak spacings between the charging steps decrease. In the figure, this occurs at Au MPCs with a radius of (roughly) 1.25 nm, corresponding to ca. 586 Au atoms10 and a capacitance of ca. 1.4 aF (butanethiolate monolayer). Effects generally similar to those in Figure 4 can be expected by varying the monolayer thickness. Increasing monolayer thickness will increase the peak spacing and more clearly reveal the quantized charging features. Experimentally, the Brust reaction4 does not yield monodisperse MPCs, but the core size dispersity can be reduced using repeated fractional recrystallization.11 We believe that voltammetric quantized capacitance charging is likely to be very sensitive to dispersity, so it is worthwhile to model its effects. A simple case is a mixture of two monodisperse clusters of different core size. We chose MPCs with mean core radii equal to those in Figure 1A; the mean radii of these MPCs are not very different. Figure 5 shows the corresponding microelectrode voltammetry simulations for mixtures of the two monodisperse MPCs in various proportions, plus derivatives of the voltammograms, which are easier to visualize. Especially instructive is the (0.5 i1 + 0.5 i2) mixture where (- ‚ ‚ -) although the charging steps can be clearly seen near EPZC, the different spacing between the peaks of the two populations dampens out the definition of the peaks at more negative and positive potentials. A more realistic simulation assumes that an MPC has a Gaussian distribution of core sizes (eq 7a), this is shown in Figure 6. The figure assumes various levels of dispersity. For a radius dispersity (σ) that is about one-third of the mean radius, the peaks furthest removed from EPZC are substantially lost. Even